FreedomGPT Generate Image

FreedomGPT Generate Image is one of the pack's "cloud generation" nodes: instead of loading a model locally, it fires your prompt off to the FreedomGPT API and brings back the generated images as a normal ComfyUI IMAGE tensor. If you're on a machine without a decent GPU - a laptop, a shared box, a rental - this is how you keep a full image pipeline running anyway, at the cost of an API key and a network round-trip.

FreedomGPT is a real third-party provider (the API base is https://chat.freedomgpt.com/api/v1), and this node is a thin client for its image endpoint. It's worth a moment of honesty: the FreedomGPT / this-pack pairing is niche enough that there's essentially no community chatter about it - you won't find a wall of forum posts. What you get instead is a straightforward API wrapper: set a key, type a prompt, and images come back as tensors you can preview, save, or feed into local post-processing. Given ComfyUI's ecosystem is heavily local-first, treat this as a "when you have no GPU or want a specific cloud model" play rather than a replacement for local sampling.

How it works

The node converts your input images (optional image through image4 ports) to PIL, builds an OpenAI-style message payload, and calls the FreedomGPT API with the Bearer key. The response - a batch of images plus a text field - comes back converted into a [N, H, W, C] tensor. Notably it supports image input, so this isn't text-to-image only: feed it a starting image and you're doing img2img or multi-image editing, depending on what the model supports. If the API returns nothing, you get a blank 512×512 tensor rather than a crash.

Inputs that matter

• api_key - paste it here, or rely on the pack's shared key file / FREEDOMGPT_API_KEY env var.
• prompt - the actual image description. system_prompt sets overall style.
• model_name - defaults to liberty; that's the current default model name, and it's worth checking FreedomGPT's docs for what's actually available before you assume it.
• number_of_images - 1–10, how many to generate.
• seed - -1 for random, or a fixed value for reproducible results.
• proxy_url - for users behind a proxy (format protocol://user:pass@addr:port, http/https/socks5 supported). Useful if you're not in a region where the API is reachable.
• timeout - seconds; 0 means no timeout.

Outputs

• images - the IMAGE batch, ready for any local node.
• text - the API's response text, in case it says something useful (like a refusal or metadata).
• history - conversation history for chaining into another call.

Install & key setup

Part of ComfyUI-YogurtNodes:

cd ComfyUI/custom_nodes
git clone https://github.com/yogurt7771/ComfyUI-YogurtNodes.git
cd ComfyUI-YogurtNodes
pip install -r requirements.txt

Restart ComfyUI and find it under "Yogurt Nodes" → LLM. Keys can be set three ways (priority order): the node's api_key field, an api_key.json file in custom_nodes/ComfyUI-YogurtNodes/yogurt_nodes/llm/ containing {"freedomgpt": "YOUR_KEY"}, or the FREEDOMGPT_API_KEY environment variable.

Common issues

• "Authentication failed" - the key isn't resolving. Check the field, the api_key.json, and the env var in that order.
• "Request times out" - set timeout to a nonzero value, or if you're in a region where the API is blocked, configure proxy_url.
• "Blank image came back" - the API returned nothing; check the text output for the actual reason.
• "The model name is wrong" - liberty is the shipped default; verify against FreedomGPT's current model list.

If you have a working local GPU, you don't need this node. If you don't, it's the difference between running the workflow and not running it at all.